selection: Relative publication probabilities of a baggr model

In baggr: Bayesian Aggregate Treatment Effects

Relative publication probabilities of a baggr model

Description

Extracts the posterior draws (or summaries) of the selection-model parameters that describe the relative publication probabilities across |z|-intervals. In Rubin summary-data models, z = tau / se; for example, selection = 1.96 estimates the relative publication probability below the conventional two-sided 5% normal threshold compared with the implicit highest-|z| interval.

Usage

selection(bg, summary = TRUE, interval = 0.95)

Arguments

bg	a baggr() model fitted with a selection-enabled likelihood (currently model = "rubin").
summary	logical; if TRUE, returns summary statistics (mean and the uncertainty interval defined by interval).
interval	uncertainty interval width (numeric between 0 and 1), used only when summary = TRUE.

Details

selection() returns draws or summaries of the selection weights only. These are relative publication probabilities, not treatment effects or corrected treatment effects. The highest-|z| interval is the reference category and is fixed to 1, so it is not returned.

Value

If summary = FALSE, a matrix of posterior draws where each column corresponds to one non-reference |z|-interval. Otherwise, a matrix with one row per interval and columns containing the mean and bounds of the uncertainty interval. The final highest-|z| interval is fixed to relative probability 1 and is therefore not returned.

baggr documentation built on June 16, 2026, 9:06 a.m.